This paper presents SURF (speeded up robust features)-based camera motion estimation assisting image registration for bronchoscope tracking. Our method for predicting the bronchoscope camera motion comprises two stages: (1) rough camera motion predication that uses SURF features and epipolar constraints to obtain inter-frame translation (up to scale) and orientation displacements and (2) performing image registration that is initialized by the estimate based on (1). The proposed method is evaluated on phantom and patient data. Experimental results from both datasets demonstrate a significant performance boost of tracking without an additional position sensor.     «

This paper presents SURF (speeded up robust features)-based camera motion estimation assisting image registration for bronchoscope tracking. Our method for predicting the bronchoscope camera motion comprises two stages: (1) rough camera motion predication that uses SURF features and epipolar constraints to obtain inter-frame translation (up to scale) and orientation displacements and (2) performing image registration that is initialized by the estimate based on (1). The proposed method is evalua...     »